AI mindset for lawyers

0:00 So, this opportunity is for me just to share what I know and maybe it will help you in doing your work. I train people across various professions, finance people, marketing, board members uh on how they can use artificial intelligence or AI in how they are doing their work. Uh I just want to check one more time. Uh um Nicholas, can you hear me? >> Yes, I can hear you. >> Thank you so much. Okay, so we we proceed.

0:23 I'll start with a very short presentation. This does 10 slides. Then immediately after that we go into the practicals of how to apply artificial intelligence in the work that you do as as lawyers. I do know that a lot of you may have doubts you meet people who have doubts about artificial intelligence and what it can do. It's not surprising. It always happens with every technology. If you look when the calculator came it faced the same type of challenges where people were questioning is it you who produced this you use the calculator but look now the calculator is with us up to now most of us can't do without the calculator the same way with with technology there will be people who are very skeptical and for good reasons but what I can assure you or what I can encourage you to do is that don't be left out because this type for technology will not wait for you.

1:35 No matter how much resistance you give, other people will proceed and still do good work. While at least you are struggling, you waste a lot of time doing things that could be facil that could be frustrated through technology. And in the same way artificial intelligence, the only good thing is that it's much more efficient than probably some most of the technology that we have received so far in the world of work. So if you've got doubts, I encourage you to maybe start looking at changing that mindset and look at how you can proceed with the with the artificial intelligence.

1:58 Uh I also know I've looked at research. I remember a few months ago when I was presenting at a conference we looked at research there is too doubt around if you go and say to someone I prepared this presentation using artificial intelligence for example you find people start doubting there's a plent of research around that it's actually becomes worse if you tell them after you have presented and that is has been proven some of the experience that are that are there but if you go back you'll find that this is the the trend across uh uh H especially when new technology comes on board.

2:47 So but you can ask us questions as we go. I do have and and Nicholas in the meeting. We are the three of us. We train on artificial intelligence. They on the more on the technical side on the IT side. So at the end when I finish my my my my presentation will then take over and just help you to understand the security issues around how to use artificial intelligence. what are the safeguards that you can use data protection things like that that's this area you will talk about that at the at the end once I finish my presentation okay so the next important part is that I'm trying to make sure this presentation okay the next important part is that um you may also have heard when when when lawyers started using artificial intelligence that some were quoting non-existent cases there are plent of cases across the globe um that you may have read and scared you as you were trying to use artificial intelligence.

3:34 But I must also emphasize that artificial intelligence is is very good depending on what platform you are using. Currently the the three platforms that seem to be doing very well. I would rank them in terms of uh usefulness in the work that you do. Cloud will be number one. Uh chat GPT work which is also just released it will be number two. uh Google Gemini and notebook LM those are the other two but they don't have some of the capacity that the first two have if you look at GPT work it has just mimicked what cloud can do so but I'll also demonstrate the two of them I'll probably not go into Gemini uh because Gemini depends on you uploading documents the other two you don't necessarily have to upload documents you can work in your own work workspace uh without necessarily uploading on docu douments onto uh onto the internet.

4:22 So it will work off your machine rather than you putting the documents on the on on the on the platform that that but I will explain as we go into into into that. So the other issues why would lawyers worry about artificial intelligence and everything they are good things the efficiency the quality the speed you spend a lot of time going through massive tons of documents reviewing uh laws to to align with what you're doing all these things take a lot of time that's why sometimes if you're a senior partner these are dedicated to junior staff but they also take a long time because they have to read that is the work that artificial intelligent can do efficiently and we will show you how you can do that without making mistakes or without experiencing any hallucinations in what in the output that you that that you produce.

5:28 Uh so you may have heard that there are six cases that never existed that were quoted. You remember that case of fictitious cases in the in the in in in America and the lawyer was sanctioned and I I read stories of many others being sanctioned for for citing fictitious cases. Yes, it the the AI can hallucinate but the level of hallucination has significantly gone down especially for paid versions. If you're doing your work, I would discourage you from using unpaid versions.

5:54 But I do know that cloud unpaid and also GPT unpaid they do very well even Gemini but there are others that are still in development. The rest of the other stuff are still in development and not as good as the the ones that I have tested. So arrest nations can can can be there but also remember large language models which basically what we are talking about as as AI now they use the they're trained try trained to predict the text in whatever you're asking so that's why they called large language models because they use language to interact with you but I think the level of sophistication has definitely improved from the start to the time we started using artificial intelligence Um, you also can see that AI the good thing is it can fabricate cases and they appear very genuine normal.

6:41 When you read it, you would love the language. You would love the way it reading but it may actually be fake. So in some cases a it doesn't tell you it doesn't know most of the time. So you need to be careful that because it won't tell you it doesn't know anything. It may it will invent and sounds very authoritative when you when you look at it and that's what we call hallucinations when it gives you non-existent stuff when it it invents things that do not exist.

7:05 If you were brainstorming for some on something it will be better but if you dealing with things that you guys handle you need to be extremely careful. Um so how do you approach AI as a lawyer? The the best way is depending on where you're sitting just treat it as a a brilliant assistant or a brilliant junior lawyer or a junior associate who carries no risk at all because there's a partner there's someone who's going to carry that risk.

7:53 So the best way AI is your assistant and you still carry the responsibility of whatever it produces. You can't give that to AI. So that is the first important point. You must treat AI whatever tool you decide to use you must you must treat it as a brilliant junior associate who carries no risk. The good thing about AI, it can draft things quickly. It can read long documents without tiring. Can structure and format things very nicely and obviously more and above that it doesn't get irritated like what you would see when you delegate work to to an individual.

8:17 But again as I indicated it can invent authority and those things can read very genuinely and it doesn't tell you when it's a it's it's not sure. So you may delegate the drafting but remember you retain the judgment. It's very safe to delegate but you need to be careful that whatever output comes of out of there you take ownership of that. Just to highlight why this is very important as well is that when you look at how artificial intelligence the experiment that have been done now you'll find that it tend to help a lot people who are average performance the the performance goes up by almost 40%.

9:04 If you look at the Boston experiment, Boston consulting experiment where they did the experiment with consultants, the performance of the of the average performer went up by 40%. And those for top performers went up by 27%. So it tend to narrow the gap between your top lawyers, I'm using this literally as an example, the top lawyers and maybe an average lawyer in terms of quality of output. But again, the ultimate judgment in terms of what comes out is entirely uh yours.

9:26 So even even after it has produced a particular document, please never cite a case you have not read in a in a in a law report or some uh um credible source. Uh if it gives you any source and quotes any law, you must be able to go and verify that that's that is correct. So every case and every stat provision comes from a law report or proper database. read before you use it. Even after AI has drafted that ask asking the tool to confirm its own citation.

9:52 Most of the time you find that it will confirm even the lawyer the New York lawyer that we were talking about who had those fictitious case laws. They even verified and asked the the the AI to to to to confirm if it's it's correct and said yes. For those that were watching CNN last night, you may also have noticed that there was someone who is suing Chad GPT or Open AI because they said CH GP gave me bad medical advice and they nearly died out of it.

10:39 And they if you look at the message that they were displaying, it was confirming that it's correct and sounding very authoritative like a like a doctor. So you need you still need to verify and we also know that these AI tools tend to verify people who know what they're doing who are very knowledgeable because if something comes out and I know it's not true or it's not correct you can always address that. Um if you're unsure in terms of confidence, if you're unsure about client data, uh don't put client data if you're unsure about uh the security and things which is going to to talk about.

11:01 But I can assure you that it's very safe uh as I'm going to take you through that especially when you're using chat GPT work GPT core work as I'll show you and also uh so chat GPT work and the cloud core work the name sounds very similar because GP just copied what cloud did and then so when you look at Zimbabwean law treat every answer that you get from this AI as a hypothesis to be tested You can't take it as as the truth and the gospel gospel.

11:47 Remember these tools the AI is basically what is called machine learning where you load all this information and it's learned from that that information. So if you look at most of the tools that are being chained out now these tools were largely trained on foreign common law jurisdiction uh not on our stats in Zimbabwe. So you may find there are gaps in terms of how it's citing there may be gaps in terms of what it's giving you uh because it doesn't have that much information but as we as people more and people more and more people interact with the law and more more and more information is put online you'll find that they tend to be much better.

12:32 So you can see there if you talking about other jurisdictions especially American European you find that most of the time very accurate very little hallucination happening there but when it comes to Zimbabwen laws there are gaps so as you use them you must understand that there will be gaps that's why I'm emphasizing that you must verify everything that comes out must be verified verification takes less time especially on factual information so you must be able to defend every line that comes out of your of your AI work whatever you're doing with the with with AI.

12:55 So I'll start I'll stop sharing this one and move to then go to show you the screen. Can you confirm Nicholas you can